Image scanners convert a visible image on a document or photograph, or an image in a transparent medium, into an electronic form suitable for copying, storing or processing by a computer. An image scanner may be a separate device or an image scanner may be a part of a copier, part of a facsimile machine, or part of a multipurpose device. Reflective image scanners typically have a controlled source of light, and light is reflected off the surface of a document, through an optics system, and onto an array of photosensitive devices. The photosensitive devices convert received light intensity into an electronic signal. Transparency image scanners pass light through a transparent image, for example a photographic positive slide, through an optics system, and then onto an array of photosensitive devices. Common photosensor technologies include Charge Coupled Devices (CCD), Charge Injection Devices (CID), Complementary-Metal-Oxide (CMOS) devices, and solar cells. Typically, for a CID or a CMOS array, each photosensitive element is addressable. In contrast, CCD line arrays commonly serially transfer all the charges, bucket-brigade style, from each line array of photosensitive elements to a small number of sense nodes for conversion of charge into a measurable voltage. The present patent document is primarily concerned with photosensor arrays having serial charge transfer registers, also called serial readout registers. In general, there is an ongoing demand for increased resolution and speed, improved color quality and image quality, and reduced cost, demands that often directly conflict and require trade-offs. The following background presents some of the factors affecting resolution, speed, color quality, image quality and cost. In general, image scanners use an optical lens system to focus an image onto an array of photosensors. Photosensor arrays typically have thousands of individual photosensitive elements. Each photosensitive element, in conjunction with the scanner optics system, measures light intensity from an effective area on the document defining a picture element (pixel) on the image being scanned. Optical sampling rate is often expressed as pixels per inch (or mm) as measured on the document (or object, or transparency) being scanned. Optical sampling rate as measured on the document being scanned is also called the input sampling rate. Photosensor assemblies for image scanners commonly have three or four line arrays of sensors, with each line array receiving a different band of wavelengths of light, for example, red, green and blue. Each line array may be filtered, or white light may be separated into different bands of wavelengths by a beam splitter. Typically, the pitch (spacing of individual photosensor elements) is the same for each line array, and typically the pitch is set to provide a specified native input sampling rate. The native input sampling rate is determined by the optics and the pitch of the individual sensors. A scanner operator may select a sampling rate that is less than the native input sampling rate by simply dropping selected pixels, or by using digital resampling techniques. Alternatively, a scanner operator may select a sampling rate that is greater than the native input sampling rate, where intermediate values are computed by interpolation. Typically, all the charges or voltages are read from the photosensor array, and are then digitized, and then subsampling or interpolation is performed on the resulting digital pixel data. Bit depth is the number of bits captured per pixel. Typically, a pixel is specified in a three-dimensional color space with a fixed number of bits in each dimension. For example, a pixel may be specified in red, green, blue (RGB) color space, with 8 bits of red information, 8 bits of green information, and 8 bits of blue information, for a total of 24 bits per pixel. Alternatively, a pixel may be specified in a cylindrical color space in which the dimensions are luminance, chrominance, and saturation. Alternatively, a three-dimensional CIE color space may be used, for example, CIELAB or CIELUV, where one dimension is luminance. In this application, “high” bit depth means that all bits are accurate, distinguishing accuracy from simple resolution. That is, a scanner could provide many bits of information, but have a noise level that makes most of the lower order bits meaningless. Even if a sensor is receiving no light, some thermal noise (called dark noise) may occur. Thermal noise (dark noise) is proportional to time. During exposure to light, the primary noise source (called shot noise) is related to conversion of photons to electrons, and the noise increases with the square root of the signal. Small sensors tend to have a lower signal-to-noise ratio than large sensors, particularly for low reflectance or low transmissivity areas of a document. Smaller sensor areas can provide higher input sampling rates, but other measures of image quality, and in particular color quality, as measured by signal-to-noise, may be reduced. If an input sampling rate is selected that is lower than the native input sampling rate, then the signal-to-noise may be improved by averaging samples. Analog signals from adjacent sensor areas may be added, or digital values may be averaged after analog-to-digital conversion. Adding N samples improves the signal-to-noise ratio by the square root of N. Typically, adding analog signals requires the signal levels to be relatively small before adding to avoid saturating a charge element, so that analog averaging is typically used for speed (fewer conversions) rather than for improvement in signal-to-noise ratio. Scanning speed is affected by multiple factors: exposure time, shift time of registers multiplied by number of pixels being shifted, output amplifier speed, and analog-to-digital conversion time. Typically, for low native input sampling rates, the primary limiter is exposure time, that is, the time required to generate a signal that provides an acceptable signal- to-noise ratio. However, if the number of pixels being shifted and converted becomes very large, then the time required to shift and convert the individual pixel signals may become the limiting factor.
Areas of an image with slowly varying color, particularly dark colors, require high bit depth and high signal-to-noise to accurately reproduce the smooth tone and texture of the original. For areas of slowly varying color, high input sampling rate is not needed because there is no high frequency information in the image. Areas of an image that change color rapidly, for example a forest scene, or a close-up photograph of a multi-colored fabric, need a high input sampling rate to capture the high frequency information, but high bit depth and high signal-to-noise are not needed. That is, for high frequency information, the color accuracy of each individual pixel is less important. High input sampling rates require small sensor areas, which in turn have relatively low signal-to-noise ratios, relatively low bit depth, and relatively low scanning speed. Large sensor areas provide high signal-to-noise, high bit depth, and high speed, but cannot provide high input sampling rates. There is a need for a scanner that provides high speed, and high color quality, and high native input sampling rate, with minimal increase in cost.